The warning of membrane fouling is of great important to maintain the stable operation of membrane bioreactor (MBR). However, traditional methods are so error-prone that probably do not acquire reliable solutions of membrane fouling due to its uncertainties. To overcome this problem, an intelligent warning method is proposed to monitor the status of MBR in this paper. The main advantages in this paper are as follows. First, an identification method, based on robust deep neural network (RDNN), is developed to diagnose the different types of membrane fouling. Second, a decision-making method, based on the restricted Boltzmann machine (RBM), is designed to distinguish the operational suggestion. Third, an intelligent warning system, based on the above two methods and some sensors, is developed to mitigate the membrane fouling in real wastewater treatment plants. Finally, the simulation and experimental results demonstrate the proposed warning method can obtain the higher identification accuracy of membrane fouling than other methods.

膜污染预警对于维持膜生物反应器（MBR）的稳定运行具有重要意义。然而，传统方法非常容易出错，由于其不确定性，可能无法获得可靠的膜污染解决方案。为了克服这个问题，本文提出了一种智能预警方法来监控MBR的状态。本文的主要优点如下。首先，开发了一种基于鲁棒深度神经网络 (RDNN) 的识别方法来诊断不同类型的膜污染。其次，设计了一种基于受限玻尔兹曼机（RBM）的决策方法来区分操作建议。三、智能预警系统，基于以上两种方式和一些传感器，开发用于减轻实际废水处理厂中的膜污染。最后，仿真和实验结果表明，所提出的预警方法能够获得比其他方法更高的膜污染识别精度。